A New Method for Quantifying Heterochrony in Evolutionary Lineages

Paleobiology, 47(2), 2021, pp. 363–384 DOI: 10.1017/pab.2020.17

Article

A new method for quantifying heterochrony in evolutionary lineages

James C. Lamsdell

Abstract.—The occupation of new environments by evolutionary lineages is frequently associated with morphological changes. This covariation of ecotype and phenotype is expected due to the process of natural selection, whereby environmental pressures lead to the proliferation of morphological variants that are a better fit for the prevailing abiotic conditions. One primary mechanism by which phenotypic variants are known to arise is through changes in the timing or duration of organismal development resulting in alterations to adult morphology, a process known as heterochrony. While numerous studies have demonstrated heterochronic trends in association with environmental gradients, few have done so within a phylogenetic context. Understanding species interrelationships is necessary to determine whether morphological change is due to heterochronic processes; however, research is hampered by the lack of a quantitative metric with which to assess the degree of heterochronic traits expressed within and among species. Here I present a new metric for quantifying heterochronic change, expressed as a heterochronic weighting, and apply it to xiphosuran chelicerates within a phylogenetic context to reveal concerted independent heterochronic trends. These trends correlate with shifts in environmental occupation from marine to nonmarine habitats, resulting in a macroevolutionary ratchet. Critically, the distribution of heterochronic weightings among species shows evidence of being influenced by both historical, phylogenetic processes and external ecological pressures. Heterochronic weighting proves to be an effective method to quantify heterochronic trends within a phylogenetic framework and is readily applicable to any group of organisms that have well-defined morphological characteristics, ontogenetic information, and resolved internal relationships.

James C. Lamsdell. Department of Geology and Geography, West Virginia University, Morgantown, West Virginia 26506, U.S.A. E-mail: [email protected]

Accepted: 15 March 2020 Data available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.pzgmsbcgp

Introduction explosion (Erwin and Valentine 2013), invasion of freshwater ecosystems (Davis et al. 2018), Understanding the evolutionary patterns colonization of land (Benton 2010), and recov- and processes that drive the generation of ery from mass extinction events (Toljagić and new phenotypes and occupation of previously Butler 2013). In turn, the innovation of pheno- unexploited environments (the occurrence of types is recognized to occur through heritable novelty and innovation, sensu Erwin 2017)is alterations in the timing of an organism’sdevel- a fundamental goal of evolutionary study. opment, a process known as heterochrony The topic of novelty and innovation has most (McNamara and McKinney 2005;McNamara frequently been explored through the lens of 2012; Colangelo et al. 2019). Therefore, hetero- adaptive radiation (Losos 2010; Yoder et al. chronic processes result in new morphological 2010), which posits that ecological opportun- characteristics that can allow organisms to exploit ities combined with chance phenotypic evolu- new environments and subsequently diversify. tion permits exploration of new ecospace Natural biological systems are complex (Stroud and Losos 2016). This model of adap- entities, in which both environmental inter- tive radiation has been invoked as the domin- action and historical components are poten- ant causal factor of major evolutionary events tially equally important (Eldredge and Salthe in Earth’s history, including the Cambrian 1984; Vrba and Eldredge 1984; Erwin 2015b;

© 2020 The Paleontological Society. This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited. 0094-8373/21 Downloaded from https://www.cambridge.org/core. IP address: 170.106.202.58, on 01 Oct 2021 at 16:50:49, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2020.17 364 JAMES C. LAMSDELL

Lamsdell et al. 2017; Congreve et al. 2018). Evo- for each species within the analysis is devel- lutionary trajectories are defined by the inter- oped. Each character represents an aspect of action of two semi-independent genealogical morphology that may exhibit a paedomorphic, and ecological hierarchies (Eldredge and Salthe peramorphic, or neutral heterochronic expres- 1984; Congreve et al. 2018); as such, ecological sion. The peramorphic and paedomorphic con- changes have the potential to influence hetero- ditions for each character are determined based chronic processes to the same extent as hetero- on a ranked series of criteria: chronic changes might alter the potential for ecological occupation. Therefore, pluralistic 1. direct observations of ontogenetic changes approaches are required to determine the rela- within the target species; tive importance of ecological or evolutionary 2. ontogenetic changes observed in closely mechanisms behind biotic responses (Lamsdell related species; et al. 2017;Tuckeretal.2017;Congreveetal. 3. ontogenetic changes observed in extant 2018), representing a synthesis of macroevolu- relatives; tionary and macroecological topics and methods. 4. comparison with outgroup juvenile morph- Heterochrony, however, has rarely been stud- ology or ontogeny for character polarity; and ied within an explicitly phylogenetic context 5. comparison with outgroup adult morph- (Bardin et al. 2017). Cranial evolution of angui- ology for character polarity. morphan lizards (Bhullar 2012) and theropod dinosaurs (Bhullar et al. 2012) has been exam- Once the polarity of the morphology of hetero- ined by placing heterochronic changes within chronic expression for each character is estab- a phylogenetic hypothesis, but no attempt to lished, characters are coded for each species incorporate ecological occupation into such and assigned a score. A paedomorphic condi- investigations has been made. These studies tion is assigned a negative (−1) score, a pera- have also focused on only a subset of morpho- morphic condition is assigned a positive (+1) logical characters that may show heterochronic score, and a neutral condition is assigned a trends, which may obscure overall patterns of score of 0. If a character cannot be coded for a heterochrony within lineages, as traits display given species, either due to it not being pre- patterns of mosaic evolution (Hopkins and Lid- served or its condition being otherwise unclear, gard 2012;Huntetal.2015), whereby individual it is not given a score and does not contribute to traits may exhibit different modes of evolution the analysis. The total number of characters within a lineage. Recognizing overall hetero- within the matrix depends on the number of chronic trends within a lineage beyond the identifiable heterochronically influenced traits trajectories of individual traits is key to under- within a lineage; as such, organisms with standing the role of heterochrony in evolution. more complex morphologies and better- Furthermore, appropriate methods to quantita- understood ontogenies will result in larger het- tively compare heterochronic trends between erochronic matrices. A larger matrix allows for lineages have been lacking until now. Here, increased granularity in the heterochronic I use phylogenetic paleoecology to analyze pat- weighting, and ideally a matrix will comprise terns in ecological affinity and heterochronic 10 or more characters; however, it is possible trends across xiphosuran chelicerates and pre- to generate a heterochronic weighting for a sent a new metric for quantitatively representing taxon that has as few as three characters coded. heterochronic changes within an evolutionary Each of the characters coded for a species in lineage. the matrix contributes to the heterochronic weighting of the species, calculated as: Data and Methods n h = i=1 i ( ) — Hwj 1 Quantifying Heterochrony. Anovel n approach for quantifying heterochrony is pro- posed herein, whereby a character matrix com- where Hw is the heterochronic weighting of prising a series of multistate characters coded species j, derived from the mean of the

Downloaded from https://www.cambridge.org/core. IP address: 170.106.202.58, on 01 Oct 2021 at 16:50:49, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/pab.2020.17 QUANTIFYING HETEROCHRONY IN XIPHOSURIDS 365

combined heterochronic scores (η)ofn charac- weights retrieved for a given clade are therefore ters, resulting in a value between 1.00 (more compared with a distribution of weight scores peramorphic) and −1.00 (more paedomorphic). over an identical topology; as such, rando- In turn, the heterochronic weighting of a given mized distributions should be generated for clade is calculated as: each topology used in an analysis. In this way, disparate analyses can determine whether clades fall within the tails of the expected het- N Hw []= j=1 j ( ) erochronic weighting distribution given their Hw k 2 N phylogenetic topology. Node- versus Tip-based Analyses.—Scores for where [Hw] is the heterochronic weighting of heterochronic weighting can be applied two clade k, derived from the mean of the combined ways, through either a node-based or a tip- heterochronic weightings (Hw)inN species, based analysis (Fig. 1). Applying heterochronic again resulting in a value between 1.00 and weighting through node-based calculations −1.00. The heterochronic weighting of a clade accurately encompasses the fact that whether a therefore represents the average of the hetero- condition for a given taxon is peramorphic, chronic weightings of its constituent species; it paedomorphic, or neutral is dependent on the is explicitly not an ancestral state reconstruc- condition in its ancestor as inferred by ancestral tion. What heterochronic weighting does state reconstruction, and so it is possible for the ensure is that all taxa within an analysis have polarity of a character to change over the evolu- a directly comparable metric, permitting ana- tionary history of a clade. As such, when apply- lysis of heterochronic trends and correlation ing node-based calculations, it is critical to with other evolutionary phenomena. While document the polarity of each character for the scale of the heterochrony metric raw values each node in the phylogeny. Heterochronic within an analysis is relative and cannot be dir- weights are calculated for nodes and are reset ectly compared between analyses, patterns and at the base of each branch of the phylogeny, timing of heterochronic shifts can be directly with individual character weights representing compared between datasets in an identical the inferred shift or transition from the ancestral manner to quantitative analyses of morpho- condition. Node-based application of hetero- space (Korn et al. 2013; Hopkins and Gerber chronic weighting more accurately reﬂects the 2017). actual process of heterochrony; however, it is To determine whether the heterochronic sensitive to sampling, as failing to sample a spe- weighting of a clade represents a concerted cies can result in an incorrect assumption of trend shift distinct from what could be expected character polarity for a given node, while large from random, nondirectional evolution, the gaps in sampling can result in unnatural stack- observed clade scores are randomized across ing of transitions at sampled nodes. The analysis the tree topology. In total, 100,000 randomiza- also becomes incredibly sensitive to phylogen- tions are performed, with the distribution etic topology, as the relative order in which of the randomized heterochronic weightings nodes occur will have a major impact on then collated into a histogram for each clade character polarity. Therefore, node-based het- to which the observed clade heterochronic erochronic weighting ideally requires a well- weighting can be directly compared. If the constrained and dated phylogeny for an evenly actual weighting score falls within either tail sampled group with good ontogenetic data. In of the distribution, the weighting is signiﬁ- practice, only a minority of clades may combine cantly different from what would be expected all of these attributes, limiting the extent to under random (nondirectional) character which heterochronic weighting can be applied change. Randomizing the heterochronic weight- from a node-based perspective—although the ings across the observed phylogeny accounts for promise of phylogenetic advances in a variety vagaries such as clade size and topological rela- of molluscan groups such as ammonoids and tionships that may impact the distribution of het- gastropods, which regularly preserve details of erochronic weighting scores. The heterochronic their ontogeny and have a good fossil record,

FIGURE 1. Example of heterochronic weightings calculated from three traits evolving across a lineage comprising taxa A–G. In the top tree, evolution of the three traits is shown with their condition (peramorphic + 1, paedomorphic −1, or neutral 0) for each species and internal node of the phylogeny shown in boxes. Transitions between character states are shown beneath each branch. The polarity of a transition is dependent on the condition of the character at the preceding node; therefore, a transition to 0 from −1 would be positive (a peramorphic transition), while a transition to 0 from + 1 would be negative (a paedomorphic transition). Node-based calculations are shown on the bottom left, where heterochronic weights are derived from the transitions leading to each node or tip species, while tip-based calculations of heterochronic weights derived from the terminal character conditions of tip species are shown on the bottom right. Both analytical variations accurately capture the overall peramorphic trend among species A and B and the paedomorphic trend from species E to G. Notably, the tip-based application of the method fails to recognize the peramorphic reversal in species F; however, tip- based heterochronic weights would recognize the peramorphic inﬂuence if this were to develop into a long-term trend. Node- and tip-based calculations of heterochronic weights are therefore both equally accurate with regard to recognizing overall trends, but node-based calculations are more precise.

marks them as potential candidates for node- weighting quantiﬁes the overall outcome of based heterochronic weighting. heterochronic events. This makes the method Alternatively, the tip-based application of less precise than its node-based articulation, heterochronic weighting provides a grand aver- as it may fail to recognize relative polarity age of heterochronic traits within observed taxa changes in characteristics, but is more broadly compared with the root character polarity. applicable to groups with uneven sampling or Rather than tracing the transition of the hetero- uncertain phylogenetic topology. Tip-based chronic event, tip-based heterochronic heterochronic weighting gives an overall

indication of peramorphic or paedomorphic (Sekiguchi et al. 1988a,b; Shuster and Sekiguchi trends over the evolutionary history of a 2003) ontogeny supplemented by detailed clade; in this way, it is similar to the method studies of the early development of the thoracic recently used by Martynov et al. (2020)to opercula and book gills (Farley 2010) and com- study paedomorphosis in nudibranchs, with pound lateral eyes (Harzsch et al. 2006). Com- the distinction that heterochronic weighting paratively few studies of the extant Asian provides a quantitative rather than qualitative species’ ontogenies exist; those few that do assessment of the degree of peramorphic or focus on Tachypleus tridentatus (Sekiguchi paedomorphic traits within a species. For the et al. 1982, 1988a,b; Sekiguchi 1988), often via present study, heterochronic weighting was direct comparison with the development of applied using the tip-based procedure, as sam- Limulus. pling of Xiphosura is uneven across their evolu- Ontogenetic data also exist for a number of tionary history, as demonstrated by the lack of extinct taxa from across the xiphosurid phyl- Silurian exemplars. This is considered an ogeny. The bellinurines Euproops danae and appropriate ﬁeld test of the method, as I sus- Euproops sp., from the Carboniferous of the pect that tip-based analyses will form the bulk United States and Germany, respectively, of any subsequent studies that adopt hetero- have been subject to detailed study of their chronic weighting. postlarval development (Schultka 2000; Haug Heterochronic Character Matrix.—A hetero- et al. 2012; Haug and Rötzer 2018b; Tashman chronic character matrix for Xiphosura was et al. 2019). Ontogenetic data have also been assembled comprising 20 characters. Characters reported from the bellinurine Alanops magniﬁ- and their peramorphic and paedomorphic con- cus, known from the Carboniferous ditions were developed following the criteria Montceau-les-Mines Konservat-Lagerstätte in detailed in “Quantifying Heterochrony”.The France, a species considered to demonstrate a characters encompass a wide variety of xipho- paedomorphically derived adult morphology suran morphology, equally divided between (Racheboeuf et al. 2002). Bellinurine embryonic traits of the prosoma (anterior tagma) and traits larvae preserved within egg clutches have also of the thoracetron (posterior tagma) (Figs. 2, 3, been reported from the Carboniferous of Russia Table 1). The resulting matrix, coded for 54 and assigned to the newly erected species xiphosurid taxa, is available in the Supplemen- Xiphosuroides khakassicus (Shpinev and Vasi- tary Material. While a number of aspects of lenko 2018), which probably represent the lar- these characters also appear in the phylogenetic vae of an existing species of Bellinurus. character matrix, several (such as body size) Outside of the bellinurines, juvenile and adult are not considered appropriate phylogenetic individuals of the paleolimulid Paleolimulus characters and are not incorporated into the kunguricus, from the Permian of Russia have phylogenetic analysis. Those heterochronic been described (Naugolnykh 2017). Subadult characters that are included comprise only a and adult stages have also been differentiated minority (5%) of the total phylogenetic charac- among the available material of the Cretaceous ters, ensuring a degree of independence tachypleine Tachypleus syriacus from the Hâqel between the two character sets. and Hjoûla Konservat-Lagerstätten of Lebanon Determination of polarity of character (Lamsdell and McKenzie 2015), while the Cret- morphs for assigning peramorphic and paedo- aceous mesolimulid Mesolimulus tafraoutensis morphic conditions was performed with refer- from the Gara Sbaa Konservat-Lagerstätte of ence to the observed ontogenetic trajectories Morocco is known from juvenile material of extinct and extant species. The most exten- (Lamsdell et al. 2020). sive work on xiphosurid development and This diversity in phylogenetic sampling of ontogeny has focused on the extant American xiphosurid ontogeny reveals a number of con- species Limulus polyphemus, with studies of sistent trends in horseshoe crab development embryonic (Scholl 1977; Sekiguchi et al. 1982; across their evolutionary history that can still Sekiguchi 1988; Shuster and Sekiguchi 2003; be observed in the ontogeny of modern species Haug and Rötzer 2018a) and postembryonic (Fig. 4). These evolutionarily conserved

FIGURE 2. Heterochronic characters coded for Xiphosura encompassing aspects of overall body size and prosomal morphology, showing paedomorphic (−1), neutral (0), and peramorphic (+1) conditions. A character unavailable for coding in a species is considered missing data (?) and does not contribute to the species score.

FIGURE 3. Heterochronic characters coded for Xiphosura encompassing aspects of thoracetron and telson morphology, showing paedomorphic (−1), neutral (0), and peramorphic (+1) conditions. A character unavailable for coding in a species is considered missing data (?) and does not contribute to the species score.

ontogenetic trends include a progressive reduc- ridge from the larval form through to the tion in expression of the prosomal ophthalmic adult, a relative decrease in the length of the

TABLE 1. Character traits used in the heterochronic character matrix, detailing the ancestral or base condition and the peramorphic and paedomorphic expression of each. Diagrammatic representations of each character are shown in Figs. 2 and 3.

Condition Trait Paedomorphic Neutral Peramorphic 1. Body size Decreased 2–5 cm Increased 2. Eye spines Expressed Vestigial projection Suppressed 3. Eye position Anterior Central Posterior 4. Eye size Reduced Moderate Enlarged 5. Ophthalmic ridge Fully expressed Laterally expressed Totally suppressed expression 6. Ophthalmic spines Elongated Vestigial Totally suppressed 7. Genal spine position Anterior to intragenal notch At carapace posterior Distally on pleural projections 8. Genal spine angle Posterior Intermediate Lateral 9. Genal spine length Reduced Half total prosomal shield length Elongated 10. Limb length Elongated To carapace margin Shortened 11. Thoracetron shape Semicircular Subquadrate Triangular 12. Thoracetron tergites Fully expressed Axially expressed Suppressed 13. Free lobe segment Separate segment Laterally free Laterally reduced 14. Axial/subaxial nodes Enlarged Subdued Suppressed 15. Epimera Elongated Length equal to width at base Reduced 16. Movable spines Reduced Length ∼2× width at base Elongated 17. Pretelson epimera Reduced Length equal to width at base Elongated 18. Opercula Reduced number, vestigial Fully expressed exopod, reduced Enlarged, increased exopod only endopod endopod expression 19. Thoracetron doublure Narrowed Width ∼40% that of thoracetron Expanded 20. Telson length Shortened Approximately half total body Elongated length

prosomal appendages, a reduction in expres- M. tafraoutensis (Lamsdell et al. 2020), P. kungur- sion of the free lobe segment, an increase in icus (Naugolnykh 2017), Paleolimulus woodae the number and complexity of opisthosomal (Lerner et al. 2016), and Vaderlimulus tricki (Ler- opercula, and an increase in the relative length ner et al. 2017). Three taxa were removed from of the telson compared with the body. It is these the matrix: Anacontium brevis, which is a syno- traits, alongside other observed consistent nym of the co-occurring Anacontium carpenteri trends in xiphosuran ontogeny, that were incor- (Anderson 1997); Liomesaspis leonardensis, porated into the heterochronic character matrix. which is poorly preserved and lacks diagnostic Phylogenetic Analysis.—Aphylogeneticcharacters separating it from Liomesaspis laevis framework for Xiphosura was constructed (see Tasch 1961); and Willwerathia laticeps, using an expanded version of the chelicerate which may not be a chelicerate (Lamsdell character matrix from Lamsdell (2016), which 2019). In total, 54 Xiphosura (sensu Lamsdell itself draws on matrices presented in Lamsdell 2013, 2016) were included, and it is these taxa (2013), Selden et al. (2015), Lamsdell and that were the focus of subsequent analyses of McKenzie (2015), and Lamsdell et al. (2015). heterochrony and ecological afﬁnity, with the The resulting matrix comprises 256 characters remaining 99 taxa within the matrix serving to coded for 153 taxa and is available in the online adequately root and demonstrate the mono- MorphoBank database (O’Leary and Kaufman phyly of the xiphosuran clade. Additionally, 2012) under the project code p2606 (accessible the character coding for Yunnanolimulus luopin- from http://morphobank.org/permalink/? gensis was updated to incorporate information P2606) as well as in the Supplementary Material. from recently described specimens exhibiting Sampling of Xiphosurida within the matrix was exceptional preservation (Hu et al. 2017). Four increased by six, incorporating Limulus wood- new characters were also added to the matrix, wardi (Watson 1909) and the newly described encompassing cardiac lobe width (character Limulitella tejraensis (Błażejowski et al. 2017), 32), the cardiac lobe bearing a median cardiac

FIGURE 4. Ontogenetic sequence of Limulus polyphemus from the Yale Peabody Museum teaching collection, beginning with the hatchling (fourth molt) and proceeding to the adult (post–22nd molt, which corresponds to the 18th posthatching molt). The ﬁnal molt is represented by specimen YPM IZ 070174. The size of each instar has been standardized to more clearly demonstrate changes in relative morphological proportions. Scale bars, 1 mm.

ridge with rounded cross section (character 38), of simulated data (Wright and Hillis 2014; the occurrence of a genal groove (character 45), O’Reilly et al. 2016; Puttick et al. 2017, 2019), and the course and extent of the genal groove although tree topologies derived from empir- (character 46). Morphological terminology for ical data analyzed under parsimony methods character deﬁnitions follows Selden and Siveter exhibit higher stratigraphic congruence than (1987) and Lamsdell (2013). those retrieved from Bayesian analysis (Sansom Tree inference was performed using Bayesian et al. 2018). Bayesian inference was performed statistical analysis, which has been shown to using Markov chain Monte Carlo analyses as outperform maximum parsimony analyses implemented in MrBayes 3.2.6 (Huelsenbeck

and Ronquist 2001), with four independent shown that horseshoe crabs invaded nonmar- runs of 100,000,000 generations and four chains ine environments multiple times over their evo- each under the maximum-likelihood model for lutionary history (Lamsdell 2016), a transition discrete morphological character data (Mkv + associated with a variety of biomechanical Γ;Lewis2001), with gamma-distributed rate and physiological challenges (Lamsdell in variation among sites. All characters were press), and it is these macroevolutionary events treated as unordered and equally weighted that comprise the focus of the work herein. As (Congreve and Lamsdell 2016). Trees were with the characteristics of Kiessling and Aber- sampled with a frequency of every 100 genera- han (2007) and Hopkins (2014), salinity is tions, resulting in 1,000,000 trees per run. The reduced to a binary variable, and species are first 25,000,000 generations (250,000 sampled assigned either a marine or nonmarine (includ- trees) of each run were discarded as burn-in, ing freshwater and deltaic/transitional and the 50% majority rule consensus tree was marginal-marine environments with a large calculated from the remaining 750,000 sampled amount of continental flora and fauna) affinity. trees across all four runs; this represents the The ecological affinity for each taxon was esti- optimal summary of phylogenetic relation- mated through the modified method of Miller ships given the available data (Holder et al. and Connolly (2001). This is a departure from 2008). Posterior probabilities were calculated Hopkins (2014), who implemented a refined from the frequency at which a clade occurred version of the Bayesian estimation of affinity among the sampled trees included in the con- developed by Simpson and Harnik (2009), sensus tree. which has traditionally been applied to higher Additionally, maximum parsimony analysis taxa (e.g., genera). The drawback of applying was performed using TNT (Goloboff et al. the Bayesian estimation method to species 2008) (made available with the sponsorship of data is that it requires multiple samples; Simp- the Willi Hennig Society) to test for topological son and Harnik (2009) only considered genera congruence between Bayesian and parsimony with at least four occurrences. This is impos- methods. The search strategy employed sible for many species in the fossil record, par- 100,000 random addition sequences with all ticularly unmineralized marine organisms characters unordered and of equal weight (Con- (which include horseshoe crabs), the majority greve and Lamsdell 2016), each followed by tree of which are known solely from single local- bisection-reconnection branch swapping (the ities—representing only a single occurrence. mult command in TNT). Jackknife (Farris et al. In contrast, the method of Miller and Connolly 1996), bootstrap (Felsenstein 1985), and Bremer (2001) simply calculates the affinity of a taxon (Bremer 1994) support values were also calcu- for a given environment as the number of lated in TNT and the ensemble Consistency, occurrences within the environment divided Retention and Rescaled Consistency Indices by the total number of occurrences, resulting were calculated in Mesquite 3.02 (Maddison in a metric ranging from 0 to 1; when com- and Maddison 2018). Bootstrapping was per- bined, the value of the metric for each of the formed with 50% resampling for 1,000 repeti- two possible affinities and no clear affinity tions, while jackknifing was performed using will always sum to 1. The advantage of this simple addition sequence and tree bisection- method is that it can operate with only a single reconnection branch swapping for 1,000 repeti- observation, with the drawback that lower tions with 33% character deletion. sample sizes increase the chances of a false Ecological Affinity.—Previous work by Kies- positive of ecological affinity. However, species sling and Aberhan (2007) and Hopkins (2014) tend to have a narrow range of habitat prefer- has set out a series of standard environmental ences (Saupe et al. 2014, 2015), thereby redu- categories for organisms found in marine envir- cing the chance of occurrences outside their onments; latitudinal occupation, substrate pref- preferred environments; when species do erence, bathymetry, and reef association. For have broader habitat preferences, they also the present study, however, a novel characteris- tend to have broader geographic ranges, tic is considered: salinity. Previous work has increasing the probability of fossilization and

thereby increasing the available sample size, ascertain the statistical significance of differ- reducing the chance of false positives. ences in heterochronic weightings across eco- The resulting metrics for ecological affinity logical affinities and phylogenetic clades. were interpreted as any single value >0.5 indi- Several analyses were performed: comparison cating a preference for that environment. between marine and nonmarine habitats, com- When no single environment scored >0.5, the parison between clades, and a test of the inter- taxon was considered to exhibit no affinity. action between clade and habitat categories. Ties (an instance where the highest-scoring For the comparison between marine and non- environment achieved a 0.5) were also consid- marine habitats, all species were included ered an indication of no affinity. In practice, in the analysis. However, for analyses that as the majority of horseshoe crab species are required the assignment of species to clades, known from only one or a handful of localities, only species assignable to Bellinurina, Paleoli- most species showed a clear ecological affinity mulidae, Austrolimulidae, or Limulidae were (see Supplementary Material). included. This resulted in the exclusion of Luna- Phylogenetic Paleoecology.—The emergent taspis aurora and Kasibelinurus amoricum (stem field of phylogenetic paleoecology leverages Xiphosurida), Bellinuroopsis rossicus and Rolfeia phylogenetic information to assess the long- fouldenensis (stem Limulina), and Valloisella lie- term evolutionary significance of ecological vinensis (stem Limulidae), as their inclusion trends within lineages through the synthesis would necessitate the formulation of unnatural, of phylogenetic theory and quantitative paleo- paraphyletic “groups” that possess no ecology (Lamsdell et al. 2017). Using the histor- evolutionary cohesion and could serve to bias ical phylogenetic and ecological framework, it analyses. Significance was estimated by permu- is possible to test for phylogenetic or ecological tation across groups with 10,000 replicates, signal in the distribution of heterochronic including Bonferroni correction to correct for weightings, thereby accounting for the possible multiple comparisons in the case of compari- influence of both the genealogical and eco- son between clades. logical biological hierarchies (Congreve et al. To test whether the distributions of species’ 2018). heterochronic weightings within clades exhibit The phylogenetic framework also permits for directional trends, Spearman’s(1904) rank cor- the estimation of the ancestral conditions of relation coefficient ρ, calculated using the cor.t- both morphological and ecological characteris- est function in R (R Core Team 2018), was used tics. To determine the number of transitions to compare the heterochronic weighting of between marine and nonmarine environments, observed taxa within a clade to their distance the ecological affinity of each species as deter- from the root of the clade. A statistically signifi- mined above was mapped onto the resulting cant ρ, in combination with a statistically sig- phylogenetic tree and parsimony-based ances- nificant clade heterochronic weighting [Hw], tral state reconstruction performed in Mesquite was considered evidence of a concerted hetero- 3.02 (Maddison and Maddison 2018). The chronic trend within the clade. The distance of a inferred ancestral ecological affinities were species from the root of the clade was deter- used to polarize the internal nodes of the tree, mined using the cladistic rank method (Gau- thereby establishing whether the ecological thier et al. 1988; Norell and Novacek 1992a,b; affinities of tip species represent inherited habi- Norell 1993; Benton and Storrs 1994; Benton tat preferences or are the result of a shift in the and Hitchin 1997; Carrano 2000, 2006; Wagner occupied niche. and Sidor 2000; Poulin 2005; Prado and Alberdi Multivariate statistical tests (permutational 2008; Huttenlocker 2014). Cladistic rank is multivariate analysis of variance [PERMA- determined by counting the sequence of pri- NOVA] using the Euclidian distance measure) mary nodes in a cladogram from the basal were performed in the statistical software pack- node to the ultimate node (Fig. 5). The main age PAST (Hammer et al. 2001) and using the axis of the cladogram is defined as the internal adonis function in the R (R Core Team 2018) branch supporting the most inclusive clade package vegan (Oksanen et al. 2019)to after each bifurcation. Smaller offshoot clades

FIGURE 5. Example of the method for assigning taxa to clade ranks for Spearman’s rank correlation.

are reduced to polytomies for the purpose of resolving as a paraphyletic grade at the base of assigning a clade rank, with each constituent Xiphosura, K. amoricum now resolves as the species within a polytomy receiving the same sister taxon to the clade comprising the two sub- rank. The assignment of equal rank to all consti- orders Limulina and Bellinurina, while ‘Kasibeli- tuents of an offshoot clade is a conservative nurus’ randalli resolves at the base of Bellinurina. approach that ensures any revealed trend is This leaves L. aurora, the sole described Ordovi- the majority pattern demonstrated by the cian species, as the sister taxon to all other most inclusive clades within the cladogram, Xiphosura. Casterolimulus, previously resolved without biasing the analysis through exclusion as the sole post-Triassic representative of of taxa located in offshoot clades. This was Austrolimulidae, now forms a clade with Victa- done for each of the four main xiphosurid limulus within Limulidae. Victalimulus and Cas- clades: Bellinurina, Paleolimulidae, Austroli- terolimulus are both Cretaceous taxa known mulidae, and Limulidae. from nonmarine environments, and the notion that they share common ancestry has been suggested previously (Lamsdell in press). The Results internal relationships of the bellinurines and Analysis of the phylogenetic character matrix limulids are also better resolved, and a sister resulted in Bayesian inference and parsimony clade to the xiphosuran crown group (composed optimality criteria retrieving a concordant tree of Tachypleinae and Limulinae) comprising topology, with the strict parsimony consensus Mesolimulus, ‘Limulitella’ vicensis, ‘Limulus’ of six most parsimonious trees having an identi- woodwardi, Victalimulus,andCasterolimulus is cal hypothesis of xiphosuran relationships to the recognized for the first time. Bayesian majority rule consensus (Fig. 6;for The distribution of environmental affinity details of the parsimony tree and the full Bayes- across the phylogeny of Xiphosura supports ian tree, see Supplementary Material). The previous assertions that the group has invaded phylogenetic topology, while broadly consistent nonmarine environments from the marine with those of previous analyses (Lamsdell 2013, realm multiple times over the course of its evo- 2016;LamsdellandMcKenzie2015), is more lutionary history (Lamsdell 2016). However, stratigraphically congruent due to movement whereas previous estimates suggested that the in the placement of K. amoricum, ‘Kasibelinurus’ transition from a marine to nonmarine environ- randalli,andCasterolimulus kletti.Previously mental affinity had occurred at least five times

FIGURE 6. Bayesian phylogeny of xiphosurids showing environmental afﬁnity of salinity and heterochronic weighting mapped onto the tree. Environmental afﬁnity is indicated on the branches (blue, marine; brown, nonmarine), heterochronic weighting is shown at the tips alongside the taxon names through heat-map shading (green, more paedomorphic; orange, more peramorphic). Bayesian posterior probabilities are shown below each node. The clades shown in Figs. 7 and 8 are labeled alongside the tree.

during xiphosuran evolution, the present results Austrolimulidae, as well as Valloisella and the indicate such a transition has occurred only four, sister taxa Victalimulus and Casterolimulus. or potentially as few as three, times (Fig. 6). While Bellinurina and the Victalimulus/ Ancestral state reconstruction demonstrates Casterolimulus clade are clearly derived from that a marine afﬁnity is plesiomorphic for marine-dwelling lineages that subsequently Xiphosura and that shifts to a nonmarine afﬁn- transitioned to nonmarine environments, the ity occurred for the clades Bellinurina and situation for Valloisella and the austrolimulids

Val- TABLE 2. One-way permutational multivariate analysis of is less clear. Due to the tree topology, with η2 — — variance (F(1,53) = 4.197, = 0.075, p = 0.0424), 10,000 loisella a nonmarine xiphosurid resolving as permutations, Euclidean distance measure. Value in sister taxon to a large clade divided between regular font is the p-value, value in italics is the raw F-value. the predominantly marine Limulidae and the Total sum of squares = 5.120, within-group sum of squares = 4.737, between-group sum of squares = 0.383. nonmarine Austrolimulidae, ancestral state reconstruction is unable to determine a clear Marine Nonmarine environmental affinity for the internal nodes Marine — 4.197 between Limulidae, Austrolimulidae, and Val- Nonmarine 0.045 — loisella. These results infer a number of potential evolutionary scenarios. One possibility is that Valloisella and Austrolimulidae independently weightings of the xiphosuran clades Bellinur- transitioned to nonmarine environments and a ina, Paleolimulidae, Austrolimulidae, and marine affinity is plesiomorphic for Limulidae. Limulidae (Table 3). Environmental affinity Alternatively, Valloisella, Austrolimulidae, and and phylogenetic relatedness therefore both Limulidae may all be derived from fully eury- influence heterochronic weightings; however, haline ancestors that genuinely have no environ- there appears to be no interaction between the mental affinity and occupied both marine and two factors, indicating that they exhibit con- nonmarine environments; in this case, the flicting signals and do not covary (Table 4). transition to a nonmarine affinity in Valloisella Comparing the heterochronic weightings of and Austrolimulidae, and a marine affinity in Bellinurina, Paleolimulidae, Limulidae, and Limulidae, would be the result of a reduction Austrolimulidae to the distribution of rando- in their ancestral occupied niche. Finally, it is mized heterochronic weightings reveals the possible that the transition to nonmarine envir- heterochronic weightings of Bellinurina, Aus- onments occurred in the ancestors of Valloisella, trolimulidae, and Limulidae to be significantly Austrolimulidae, and Limulidae and that a non- different from what would be expected from marine affinity is therefore plesiomorphic for random, while that of Paleolimulidae falls all three taxa, with Limulidae having undergone within what could be explained from a random a subsequent reversal to reoccupy marine distribution (Fig. 7). Interestingly, Spearman’s environments. rank correlation (Fig. 8), indicates directional Heterochronic weightings are unevenly trends in heterochronic weighting within Aus- distributed across the phylogeny, with more trolimulidae and Bellinurina, but no significant extreme (i.e., higher positive or negative) trend within Limulidae or Paleolimulidae. The values occurring among taxa with a nonmarine heterochronic weightings of Bellinurina and environmental affinity. This distinction is borne Austrolimulidae show a consistent directional out statistically, with PERMANOVA tests trend as demonstrated by locally estimated showing that the variance of heterochronic scatterplot smoothing (LOESS) regression weightings of all species occupying marine (although Bellinurina do undergo a slight environments is significantly distinct from shift in trajectory among their highest-ranked that of those occupying nonmarine environ- clades), while Paleolimulidae and Limulidae ments (Table 2). Statistically significant dif- exhibit random directional shifts across the ferences also separate the heterochronic phylogeny. This distinction is potentially

η2 TABLE 3. One-way permutational multivariate analysis of variance (F(3,49) = 73.87, = 0.83, p = 0.0001) excluding stem taxa, 10,000 permutations, Euclidean distance measure. Values in regular font are Bonferroni corrected p-values, those in italics are raw F-values. Total sum of squares = 5.088, within-group sum of squares = 0.8588, between-group sum of squares = 4.2292.

Bellinurina Paleolimulidae Austrolimulidae Limulidae Bellinurina — 14.79 113.5 175.4 Paleolimulidae 0.0024 — 16.94 20.44 Austrolimulidae 0.0006 0.0204 — 20.79 Limulidae 0.0006 0.0036 0.0018 —

TABLE 4. Two-way permutational multivariate analysis of In comparison, the observed heterochronic variance excluding stem taxa, 10,000 permutations, Euclidean distance measure. weighting for Limulidae falls within the tail of the randomized distribution, with a value Source of Sum of Mean equal to that of a set of those retrieved from variation squares df squares Fp the randomizations. Habitat 0.46075 1 0.46075 10.155 0.0001 Clade 3.77600 3 1.25867 27.742 0.0001 Interaction 0.00005 3 <0.00002 <0.0004 1 Discussion Residual 1.86020 41 0.04537 Total 5.08770 48 Shifts in ecological afﬁnity correlate with changes in evolutionary regime in Xiphosura. Clades that invade nonmarine environments borne out in the position of the observed exhibit distinct differences in the prevalence of heterochronic weightings relative to the rando- heterochronic traits in comparison to those that mized distribution of heterochronic weight- inhabit the marine realm (Fig. 6), with Austroli- ings: the observed weightings for Bellinurina mulidae demonstrating increased prevalence of and Austrolimulidae sit far outside the rando- peramorphy, while paedomorphy is prevalent mized distribution, expressing a value that among Bellinurina (Fig. 7). Paedomorphic traits does not occur within the randomized weights. have long been recognized in bellinurines

FIGURE 7. Histograms showing the distribution of randomized heterochronic weightings across 100,000 permutations for Bellinurina, Paleolimulidae, Limulidae, and Austrolimulidae. The actual heterochronic weightings of the clades, derived using characters shown in Figs. 2 and 3, are indicated by the black arrows. Weightings in either tail of the distribution are considered to be more extreme than would be expected from random. The negative tail indicates the occurrence of paedomorphosis, the positive tail indicates peramorphosis.

FIGURE 8. Graphs showing distribution of heterochronic weightings along clade rank for Bellinurina, Paleolimulidae, Limulidae, and Austrolimulidae. Each plot displays a solid linear regression line and a dashed LOESS regression line. A negative slope indicates a general paedomorphic trend, while a positive slope is representative of a peramorphic trend. The results of Spearman’s rank correlation, both in terms of statistical signiﬁcance and raw ρ, are shown at the top of each graph.

(Haug et al. 2012; Lamsdell in press), including peramorphic (in Austrolimulidae) traits. It their retention of long, gracile prosomal appen- seems that shifts in environmental occupation dages into adulthood; visible opisthosomal set these lineages along a heterochronic trajec- segmentation; and elongated dorsal prosomal tory resulting in a directional bias (Gould shield spines. Austrolimulids, meanwhile, 1982), whereby changes in the timing or rate of develop elongate and splayed prosomal genal development produce innovative morphologies, spines; reduce the size of their opisthosomal ter- the selection of which—mediated by the envir- gopleura; and exhibit enlarged, posteriorly posi- onment—result in increasingly specialized phe- tioned lateral eyes—all of which are recognized notypes. Heterochronic processes are known to as peramorphic conditions herein. Interestingly, be one of the primary ways by which morpho- lineages that make the transition to nonmarine logical innovation (sensu Erwin 2015a, 2017) environments demonstrate concerted and occurs (McNamara and McKinney 2005;McNa- enduring heterochronic trends (Fig. 8)thatper- mara 2012; Colangelo et al. 2019), and it is the sist for millions of years, with species progres- nonmarine xiphosurans that contribute the sively exhibiting an increasingly greater most to xiphosuran disparity and occupy novel number of paedomorphic (in Bellinurina) or regions of morphospace (Lamsdell 2016).

It is notable that only rarely do either of the mass extinction by occupying nonmarine heterochronic trends observed here show any environments and returned to the marine indication of heterochronic reversals and that realm during the subsequent recovery, a possi- in both cases the trend is associated with a bility first suggested by Błażejowski et al. marked shift in diversity dynamics, with Belli- (2017). Second, Limulidae display on average nurina increasing in diversity before rapidly more positive heterochronic weightings than going extinct and Austrolimulidae exhibiting other marine taxa (Fig. 7), although these pera- decreased rates of speciation and a lower diver- morphic traits do not manifest as part of a con- sity even in relation to other Xiphosura (Lams- certed trend (Fig. 8) as they do in the limulid dell 2016, in press). The combination of limited nonmarine sister group, Austrolimulidae, reversal and shift in diversity dynamics sug- which is characterized by extreme peramorph- gests that these directional trends may ism. This opens up the possibility that, if limu- represent macroevolutionary ratchets (trends lids have reoccupied marine environments where reversals are rare and ultimately result from an ancestral nonmarine habitat, the ele- in increased extinction risk or a decrease in vated occurrence of peramorphic character con- rates of origination: Van Valkenburgh 1991, ditions in the clade may be a relict of the 1999; Van Valkenburgh et al. 2004), although peramorphic trajectory that continued in aus- whether this is due to consistent environmental trolimulids. If this were to be the case, it pressure or some inherent property of the would suggest that returning to marine envir- developmental processes operating is unclear. onments stopped the heterochronic bias in Numerous consistent peramorphic and paedo- limulids, and perhaps most interestingly, that morphic evolutionary trends associated with the changes the lineage had undergone while environmental gradients—termed peramor- occupying nonmarine environments were not phoclines and paedomorphoclines (McNamara subsequently reversed. 1982), or more generally heteroclines (McKin- The exact mechanism by which heterochrony ney 1999), although these terms conflate pro- operated in the cases observed here is uncer- cess (heterochrony) and evolutionary outcome tain. It is notoriously difficult (arguably impos- (directional bias or macroevolutionary ratchet- sible) to discern between changes in timing and ing)—have been recognized in the fossil record changes in rate of development without (McNamara 1982, 1986, 1988; Simms 1988; detailed ontogenetic sequences of consecutive Korn 1995; Crônier et al. 1998; McKinney species (Gould 1988; Jones 1988; McKinney 1999; Poty 2010; Fernandez-Lopez and Pavia 1988; Allmon 1994; McNamara and McKinney 2015). Most described heterochronic trends do 2005; Bardin et al. 2017). Nevertheless, it is pos- not exhibit heterochronic reversals, with only sible to recognize broad peramorphic and a few notable exceptions (Gerber 2011), paedomorphic trends. Why these processes although it should be noted that none of these unfolded in a ratchet-like fashion is also studies were performed within a phylogenetic unclear. Classical natural selection (Darwin framework. It has previously been suggested 1859) may explain the evolution of increasingly that concerted heterochronic trends are always unusual morphologies; however, it is unclear controlled by environmental factors (McKinney whether these morphologies are truly specia- 1986, 1988); however, subsequent studies have lized or simply bizarre. Another possibility documented heterochronic trends occurring among the Bellinurina is that physiological apparently independently of any environmen- changes required in order for xiphosurans to tal gradient (Breton 1997). In the present tolerate low-salinity environments for extended study, xiphosurans possibly undergo one periods of time may have been accomplished reversal in environmental affinity, with the through the retention of larval physiology ancestors of the predominantly marine Limuli- into adulthood (Lamsdell in press). The larvae dae potentially having a nonmarine or mixed of modern horseshoe crabs are extremely toler- environmental affinity (Fig. 6). The ramifica- ant of salinities lower than 35‰ (Shuster 1982; tions of this are twofold. First, it would suggest Ehlinger and Tankersley 2007; Botton et al. that xiphosurans survived the end-Permian 2010), and paedomorphic processes permitting

the maintenance of this tolerance in adults may shows evidence of being influenced by both have also resulted in the retention of larval historical, phylogenetic processes and external morphological characteristics. The broad, shal- ecological pressures. This is most clearly demon- low prosomal carapaces of austrolimulids, strated by the manner in which the independent meanwhile, may have developed in response occupation of nonmarine environments is to unidirectional hydrodynamic environments accompanied by significant heterochronic trends that the group encountered as it radiated in in both Bellinurina and Austrolimulidae, but lacustrine environments. It is worth consider- manifesting as a paedomorphic trend among bel- ing that macroevolutionary ratchets in even linurines and a peramorphic trend within austro- closely related groups may occur through dis- limulids. Therefore, while the environment can tinct mechanisms. exert a strong pressure on both phenotype (as Ultimately, one of the more interesting out- expected by the fundamental evolutionary pro- comes of the study is the support for the cess of natural selection) and the underlying quasi-independence of the signal imparted by developmental processes that govern phenotype, history and ecology in evolution. History (as it can only do so utilizing the available morpho- represented by phylogeny) and ecology are logical and developmental frameworks. The both significant sources of variation among het- availability and composition of these frame- erochronic weights but do not interact (Table 4), works is mediated by contingent, historical fac- indicating that although they both exert influence tors (as expressed by phylogeny) that limit both on the distribution of heterochronic weights, they the potential for adaptation to certain environ- do so with conflicting signals. This conflict is due mental conditions and the structural or develop- to the impact of the quasi-independent genea- mental outcome of concerted selective pressure. logical and ecological biological hierarchies (Con- A well-known example of the former is the fact greve et al. 2018), whereby historical contingency that vertebrates returning to fully aquatic envir- limits the morphological framework for subse- onments are constrained to breathing air, while quent adaptation to ecological pressures (Gould a structural example of the latter are the hook-like and Lewontin 1979; Eldredge and Salthe 1984; projections of the beak’s tomia in mergansers Anderson and Allmon 2018). Logically, and as that are used to aid in gripping their fish prey. hinted at by the data discussed here, this tension The serrations perform a function similar to between competing hierarchies can also extend that of the narrow, curved teeth of gars, which to developmental frameworks as the mediating also prey upon small fish. Teeth, however, were factors by which morphologic phenotypes are lost in the avian stem-lineage; lacking the devel- expressed. opmental framework to express teeth, mergansers instead developed modifications to the serrated tomia prevalent in ducks and geese. Conclusions Xiphosurans demonstrate that such contingent Applying this new method for quantifying processes can also affect the mechanisms by heterochrony, expressed as a heterochronic which developmental shifts occur. weighting, within a phylogenetic context reveals Heterochronic weighting has the proven concerted independent heterochronic trends in potential to be an effective method to quantify xiphosurans. These trends correlate with, and heterochronic trends within a phylogenetic may be driven by, shifts in environmental occu- framework. Comparing the observed hetero- pation from marine to nonmarine habitats, chronic weightings of clades to randomized resulting in a macroevolutionary ratchet distributions permits the discrimination of con- whereby environmental selective factors result certed heterochronic trends from what would in the preferential retention of phenotypes be expected under random (nondirectional) derived from heterochronic processes, which in character change. The method is readily applic- turn reinforces directional heterochronic trends able to any group of organisms that have well- and the proliferation of peramorphic or paedo- defined morphological characteristics, onto- morphic characteristics. Critically, the distribu- genetic information, and resolved internal relation of heterochronic weightings among species tionships; indeed, in order to test the generality

of the observations made for xiphosurans, it is Anderson, L. I. 1997. The xiphosuran Liomesaspis from the Montceau-les-Mines Konservat-Lagerstätte, Massif Central, imperative that additional studies be per- France. Neues Jahrbuch für Geologie und Paläontologie— formed in disparate clades. Future work should Abhandlungen 204:415–436. also aim to apply node-based heterochronic Bardin, J., I. Rouget, and F. Cecca. 2017. Ontogenetic data analyzed as such in phylogenies. Systematic Biology 66:23–37. weighting in appropriate groups and seek to Benton M. J. 2010. The origins of modern biodiversity on land. apply both tip- and node-based calculations Philosophical Transactions of the Royal Society of London B to the same data to further explore the behavior 365:3667–3679. Benton, M. J., and R. Hitchin. 1997. Congruence between phylogen- and comparability of both. The combination of etic and stratigraphic data on the history of life. Proceedings of the this heterochronic metric with ecological afﬁn- Royal Society of London B 264:885–890. ity data affords easy study of the correlation Benton, M. J., and G. W. Storrs. 1994. Testing the quality of the fossil record: paleontological knowledge is improving. Geology between developmental changes and environ- 22:111–114. mental shifts as a branch of phylogenetic paleo- Bhullar, B.-A. S. 2012. A phylogenetic approach to ontogeny and ecology and has the potential to open new heterochrony in the fossil record: cranial evolution and development in anguimorphan lizards (Reptilia: Squamata). Journal of avenues into studying the relationship between Experimental Zoology B 318B:521–530. evolutionary developmental processes and Bhullar, B.-A. S., J. Marugán-Lobón, F. Racimo, G. S. Bever, T. external environmental causal factors. B. Rowe, M. A. Norrell, and A. Abzhanov. 2012. Birds have paedomorphic dinosaur skulls. Nature 487:223–226. Błażejowski, B., G. Niedźwiedzki, K. Boukhalfa, and M. Soussi. 2017. Limulitella tejraensis, a new species of limulid (Chelicerata, Acknowledgments Xiphosura) from the Middle Triassic of southern Tunisia (Saharan – I thank E. Lazo-Wasem (Yale Peabody Platform). Journal of Paleontology 91:960 967. Botton, M. L., R. A. Tankersley, and R. E. Loveland. 2010. Develop- Museum) for providing photographs of Limulus mental ecology of the American horseshoe crab Limulus polyphe- instars and A. Downey (West Virginia Univer- mus. Current Zoology 56:550–562. sity) for coding assistance that aided in collating Bremer, K. 1994. Branch support and tree stability. Cladistics 10:295–304. randomization results. The concepts and per- Breton, G. 1997. Patterns and processes of heterochrony in Meso- spectives presented in this paper have been zoic goniasterid sea-stars. Lethaia 30:135–144. reﬁned over the years through in-depth discus- Carrano, M. T. 2000. Homoplasy and the evolution of dinosaur locomotion. Paleobiology 26:489–512. sion with my colleagues B. Anderson (West Vir- Carrano, M. T. 2006. Body-size evolution in the Dinosauria. Pp. ginia University), C. Congreve (North Carolina 225–268 in M. T. Carrano, T. J. Gaudin, R. W. Blob, and J. State University), A. Falk (Centre College), A. R. Wible, eds. Amniote paleobiology. University of Chicago Press, Chicago. Manafzadeh (Brown University), and J. Miya- Colangelo, P., D. Ventura, P. Piras, J. P. G. Bonaiuti, and mae (Yale University). I am especially grateful G. Ardizzone. 2019. Are developmental shifts the main driver to A. Whitaker (University of Kansas) who proof phenotypic evolution in Diplodus spp. (Perciformes: Sparidae)? BMC Evolutionary Biology 19:106. vided graphical design servicesto aid in the pres- Congreve, C. R., and J. C. Lamsdell. 2016. Implied weighting and entation of this research at the Geological Society its utility in palaeontological datasets: a study using modelled – of America annual conference and possesses phylogenetic matrices. Palaeontology 59:447 462. Congreve, C. R., A. R. Falk, and J. C. Lamsdell. 2018. Biological invaluable skills in helping to reassemble 3D hierarchies and the nature of extinction. Biological Reviews printed turtle skulls. I thank D. Bapst (Texas 93:811–826. A&M University) and an anonymous referee Crônier, C., S. Renaud, R. Feist, and J.-C. Auffray. 1998. Ontogeny of Trimerocephalus lelievrei (Trilobita, Phacopida), a representative for their detailed and thoughtful reviews that of the Late Devonian phacopine paedomorphocline: a morpho- greatly improved the article and prompted me metric approach. Paleobiology 24:359–370. to clarify aspects of the method, as well as Darwin, C. R. 1859. On the origin of species by means of natural selection; or the preservation of favoured races in the struggle encouraging me to devote further page space to for life. Bantam Books, New York. explanations of macroevolutionary phenomena. Davis, K. E., S. De Grave, C. Delmer, and M. A. Wills. 2018. Fresh- water transitions and symbioses shaped the evolution and extant diversity of caridean shrimps. Communications Biology Literature Cited 1:16. Ehlinger, G. S., and R. A. Tankersley. 2007. Reproductive ecology of Allmon, W. D. 1994. Patterns and processes of heterochrony in the American horseshoe crab Limulus polyphemus in the Indian Lower Tertiary turritelline gastropods, U.S. Gulf and Atlantic River Lagoon: an overview. Florida Scientist 70:449–463. Coastal plains. Journal of Paleontology 68:80–95. Eldredge, N. I., and S. N. Salthe. 1984. Hierarchy and evolution. Anderson, B. M., and W. D. Allmon. 2018. When domes are span- Oxford Surveys in Evolutionary Biology 1:184–208. drels: on septation in turritellids (Cerithioidea) and other gastro- Erwin, D. H. 2015a. Novelty and innovation in the history of life. pods. Paleobiology 44:444–459. Current Biology 25:R930–R940.

Erwin, D. H. 2015b. Was the Ediacaran–Cambrian radiation a Proceedings of the National Academy of Sciences USA 109:20520– unique evolutionary event? Paleobiology 41:1–15. 20525. Erwin, D. H. 2017. The topology of evolutionary novelty and Hu, S., Q. Zhang, R. M. Feldmann, M. J. Benton, C. E. Schweitzer, innovation in macroevolution. Philosophical Transactions of the W. Wen, C. Zhou, T. Xie, T. Lü, and S. Hong. 2017. Exceptional Royal Society of London B 372:20160422. appendage and soft-tissue preservation in a Middle Triassic Erwin, D. H., and J. W. Valentine. 2013. The Cambrian explosion: horseshoe crab from SW China. Scientific Reports 7:14112. the construction of animal biodiversity. Roberts and Company, Huelsenbeck, J. P., and F. Ronquist. 2001. MRBAYES: Bayesian Greenwood, Colo. inference of phylogenetic trees. Bioinformatics 17:754–755. Farley, R. D. 2010. Book gill development in embryos and first and Hunt, G., M. J. Hopkins, and S. Lidgard. 2015. Simple versus com- second instars of the horseshoe crab Limulus polyphemus L. (Chelicer- plex models of trait evolution and stasis as a response to environ- ata, Xiphosura). Arthropod Structure & Development 39:369–381. mental change. Proceedings of the National Academy of Sciences Farris, J. S., V. A. Albert, M. Källersjö, D. Lipscomb, and A. G. Kluge. USA 112:4885–4890. 1996. Parsimony jackknifing outperforms neighbor-joining. Cla- Huttenlocker, A. K. 2014. Body size reductions in nonmammalian distics 12:99–124. eutheriodont therapsids (Synapsida) during the end-Permian Felsenstein, J. 1985. Confidence limits on phylogenies: an approach mass extinction. PLoS ONE 9:e87553. using the bootstrap. Evolution 39:783–791. Jones, D. S. 1988. Sclerochronology and the size versus age prob- Fernandez-Lopez, S. R., and G. Pavia. 2015. Mollistephaninae and lem. Pp. 93–108 in M. L. McKinney, ed. Heterochrony in evolu- Frebolditinae, new subfamilies of Middle Jurassic stephanocera- tion: a multidisciplinary approach. Plenum, New York. tid Ammonoidea. Paläontologische Zeitschrift 89:707–727. Kiessling, W., and M. Aberhan. 2007. Environmental determinants Gauthier, J., A. G. Kluge, and T. Rowe. 1988. Amniote phylogeny of marine benthic biodiversity dynamics through Triassic–Juras- and the importance of fossils. Cladistics 4:105–209. sic time. Paleobiology 33:414–434. Gerber, S. 2011. Comparing the differential filling of morphospace Korn, D. 1995. Paedomorphosis of ammonoids as a result of seale- and allometric space through time: the morphological and devel- vel fluctuations in the Late Devonian Wocklumeria Stufe. Lethaia opmental dynamics of Early Jurassic ammonoids. Paleobiology 28:155–165. 37:369–382. Korn, D., M. J. Hopkins, and S. A. Walton. 2013. Extinction space—a Goloboff, P. A., J. A. Farris, and K. C. Nixon. 2008. TNT, a free pro- method for the quantification and classification of changes in gram for phylogenetic analysis. Cladistics 24:774–786. morphospace across extinction boundaries. Evolution 67:2795– Gould, S. J. 1982. The meaning of punctuated equilibrium and its 2810. role in validating a hierarchical approach to macroevolution. Lamsdell, J. C. 2013. Revised systematics of Palaeozoic “horseshoe Pp. 83–104 in R. Milkman, ed. Perspectives on evolution. Sinauer, crabs” and the myth of monophyletic Xiphosura. Zoological Jour- Sunderland, Mass. nal of the Linnean Society 167:1–27. Gould, S. J. 1988. The uses of heterochrony. Pp. 1–13 in M. L, Lamsdell, J. C. 2016. Horseshoe crab phylogeny and independent McKinney, ed. Heterochrony in evolution: a multidisciplinary colonisations of freshwater: ecological invasion as a driver for approach. Plenum, New York. morphological innovation. Palaeontology 59:181–194. Gould, S. J., and R. C. Lewontin. 1979. The spandrels of San Marco Lamsdell, J. C. 2019. A chasmataspidid affinity for the putative and the Panglossian paradigm: a critique of the adaptationist pro- xiphosuran Kiaeria Størmer, 1934. Paläontologische Zeitschrift. gramme. Proceedings of the Royal Society of London B 205:581–598. doi: 10.1007/s12542-019-00493-8. Hammer, Ø., D. A. T. Harper, and P. D. Ryan. 2001. PAST: Pale- Lamsdell, J. C. In press. Evolutionary history of the dynamic horse- ontological Statistics software package for education and data shoe crab. International Wader Studies 21. analysis. Palaeontologia Electronica 4(1). Lamsdell, J. C., and S. C. McKenzie. 2015. Tachypleus syriacus Harzsch, S., K. Vilpoux, D. C. Blackburn, D. Platchetzki, N. (Woodward)—a sexually dimorphic Cretaceous crown limulid L. Brown, R. Melzer, K. E. Kempler, and B. A. Battelle. 2006. Evo- reveals underestimated horseshoe crab divergence times. Organ- lution of arthropod visual systems: development of the eyes and isms Diversity & Evolution 15:681–693. central visual pathways in the horseshoe crab Limulus polyphemus Lamsdell, J. C., D. E. G. Briggs, H. P. Liu, B. J. Witzke, and R. Linnaeus, 1758 (Chelicerata, Xiphosura). Developmental Dynam- M. McKay. 2015. A new Ordovician arthropod from the Winne- ics 235:2641–2655. shiek Lagerstätte of Iowa (USA) reveals the ground plan of euryp- Haug, C., and M. A. I. N. Rötzer. 2018a. The ontogeny of Limulus terids and chasmataspidids. Science of Nature 102:63. polyphemus (Xiphosura s. str., Euchelicerata) revised: looking Lamsdell, J. C., C. R. Congreve, M. J. Hopkins, A. Z. Krug, and M. “under the skin.” Development Genes and Evolution 228:49–61. E. Patzkowsky. 2017. Phylogenetic paleoecology: tree-thinking Haug, C., and M. A. I. N. Rötzer. 2018b. The ontogeny of the 300 and ecology in deep time. Trends in Ecology and Evolution million year old xiphosuran Euproops danae (Euchelicerata) and 32:452–463. implications for resolving the Euproops species complex. Develop- Lamsdell, J. C., J. N. Tashman, G. Pasini, and A. Garassino. 2020. A ment Genes and Evolution 228:63–74. new limulid (Chelicerata, Xiphosurida) from the Late Cretaceous Haug, C., P. Van Roy, A. Leipner, P. Funch, D. M. Rudkin, (Cenomanian–Turonian) of Gara Sbaa, southeast Morocco. Cret- L. Schöllmann, and J. T. Haug. 2012. A holomorph approach to aceous Research 106:104230. xiphosuran evolution—a case study on the ontogeny of Euproops. Lerner, A. J., S. G. Lucas, and C. F. Mansky. 2016. The earliest paleo- Development Genes and Evolution 222:253–268. limulid and its attributed ichnofossils from the Lower Mississippian Holder, M. T., J. Sukumaran, and P. O. Lewis. 2008. A justification (Tournasian) Horton Bluff Formation of Blue Beach, Nova Scotia, for reporting the majority-rule consensus tree in Bayesian phylo- Canada. Neues Jahrbuch für Geologie und Paläontologie—Abhan- genetics. Systematic Biology 57:814–821. dlungen 280:193–214. Hopkins, M. J. 2014. The environmental structure of trilobite mor- Lerner, A. J., S. G. Lucas, and M. Lockley. 2017. First fossil horse- phological disparity. Paleobiology 40:352–373. shoe crab (Xiphosurida) from the Triassic of North America. Hopkins, M. J., and S. Gerber. 2017. Morphological disparity. Pp. Neues Jahrbuch für Geologie und Paläontologie—Abhandlun- 1–11 in L. Nuño de la Rosa and G. B. Müller, eds. Evolutionary gen 286:289–302. developmental biology. Springer, Cham, Switzerland. Lewis, P. O. 2001. A likelihood approach to estimating phyl- Hopkins, M. J., and S. Lidgard. 2012. Evolutionary mode routinely ogeny from discrete morphological character data. Systematic varies among morphological traits within fossil species lineages. Biology 50:913–925.

Losos, J. B. 2010. Adaptive radiation, ecological opportunity, and of phenotype data. Proceedings of the Royal Society of London B evolutionary determinism. American Naturalist 175:623–639. 284:20162290. Maddison, W. P., and D. R. Maddison. 2018. Mesquite: a modular Puttick, M. N., J. E. O’Reilly, D. Pisani, and P. C. J. Donoghue. 2019. system for evolutionary analysis, version 3.51. http://www.mes- Probabilistic methods outperform parsimony in the phylogenetic quiteproject.org, accessed 10 August 2019. analysis of data simulated without a probabilistic method. Palae- Martynov, A., K. Lundin, B. Picton, K. Fletcher, K. Malmberg, and ontology 62:1–17. T. Korshunova. 2020. Multiple paedomorphic lineages of soft- Racheboeuf, P. R., J. Vannier, and L. I. Anderson. 2002. A new substrate burrowing invertebrates: parallels in the origin of Xeno- three-dimensionally preserved xiphosuran chelicerate from the cratena and Xenoturbella. PLoS ONE 15:e0227173. Montceau-les-Mines Lagerstätte (Carboniferous, France). Palae- McKinney, M. L. 1986. Ecological causation of heterochrony: a test ontology 45:125–147. and implications for evolutionary theory. Paleobiology 12:282– R Core Team. 2018. R: a language and environment for statistical 289. computing. R Foundation for Statistical Computing, Vienna, McKinney, M. L. 1988. Classifying heterochrony. Allometry, size Austria. http://www.r-project.org. and time. Pp. 17–34 in M. L. McKinney, ed. Heterochrony in evo- Sansom, R. S., P. G. Choate, J. N. Keating, and E. Randle. 2018. Par- lution: a multidisciplinary approach. Plenum, New York. simony, not Bayesian analysis, recovers more stratigraphically McKinney, M. L. 1999. Heterochrony: beyond words. Paleobiology congruent phylogenetic trees. Biology Letters 14:20180263. 25:149–153. Saupe, E. E., J. R. Hendricks, R. W. Portell, H. J. Dowsett, McNamara, K. J. 1982. Heterochrony and phylogenetic trends. A. Haywood, S. J. Hunter, and B. S. Lieberman. 2014. Macroevo- Paleobiology 8:130–142. lutionary consequences of profound climate change on niche evo- McNamara, K. J. 1986. The role of heterochrony in the evolution of lution in marine molluscs over the past three million years. Cambrian trilobites. Biological Reviews 61:121–156. Proceedings of the Royal Society of London B 281:20141995. McNamara, K. J. 1988. Patterns of heterochrony in the fossil record. Saupe, E. E., H. Qiao, J. R. Hendricks, R. W. Portell, S. J. Hunter, Trends in Ecology and Evolution 3:176–180. J. Soberón, and B. S. Lieberman. 2015. Niche breadth and geo- McNamara, K. J. 2012. Heterochrony: the evolution of develop- graphic range size as determinants of species survival on geological ment. Evolution: Education and Outreach 5:203–218. time scales. Global Ecology and Biogeography 24:1159–1169. McNamara, K. J., and M. L. McKinney. 2005. Heterochrony, dispar- Scholl, G. 1977. Beiträge zur Embryonalentwicklung von Limulus ity, and macroevolution. Paleobiology 31:17–26. polyphemus L. (Chelicerata, Xiphosura). Zoomorphologie 86:99– Miller, A. I., and S. R. Connolly. 2001. Substrate affinities of 154. higher taxa and the Ordovician radiation. Paleobiology Schultka, S. 2000. Zur palökologie der Euproopiden im Nordwest- 27:768–778. deutschen Oberkarbon. Mitteilungen aus dem Museum für Nat- Naugolnykh, S. V. 2017. Lower Kungurian shallow-water lagoon urkunde in Berlin, Geowissenschaftliche Reihe 3:87–98. biota of Middle Cis-Urals, Russia: towards paleoecological recon- Sekiguchi, K. 1988. Embryonic development. Pp. 145–181 in struction. Global Geology 20:1–13. K. Sekiguchi, ed. Biology of horseshoe crabs. Science House, Norell, M. A. 1993. Tree-based approaches to understanding his- Tokyo. tory: comments on ranks, rules, and the quality of the fossil Sekiguchi, K., Y. Yamamichi, and J. D. Costlow. 1982. Horseshoe record. American Journal of Science 293A:407–417. crab developmental studies I. Normal embryonic development Norell, M. A., and M. J. Novacek. 1992a. Congruence between of Limulus polyphemus compared with Tachypleus tridentatus. Pp. superpositional and phylogenetic patterns: comparing cladistic 53–73 in J. Bonaventura, C. Bonaventura, and S. Tesh, eds. Physi- patterns with fossil records. Cladistics 8:319–337. ology and biology of horseshoe crabs: studies on normal and Norell, M. A., and M. J. Novacek. 1992b. The fossil record and evo- environmentally stressed animals. Liss, New York. lution: comparing cladistic and paleontologic evidence for verte- Sekiguchi, K., H. Seshimo, and H. Sugita. 1988a. Post-embryonic brate history. Science 255:1690–1693. development. Pp. 181–195 in K. Sekiguchi, ed. Biology of horse- Oksanen, J., F. G. Blanchet, M. Friendly, R. Kindt, P. Legendre, shoe crabs. Science House, Tokyo. D. McGlinn, P. R. Michin, R. B. O’Hara, G. L. Simpson, Sekiguchi, K., H. Seshimo, and H. Sugita. 1988b. Post-embryonic P. Solymos, M. H. H. Stevens, E. Szoecs, and H. Wagner. 2019. development of the horseshoe crab. Biological Bulletin 174:337– Package ‘vegan’. https://cran.r-project.org/web/packages/ 345. vegan/index.html, accessed 12 February 2020. Selden, P. A., and D. J. Siveter. 1987. The origin of the limuloids. O’Leary, M. A., and S. G. Kaufman. 2012. MorphoBank 3.0: web Lethaia 20:383–392. application for morphological phylogenetics and taxonomy. Selden, P. A., J. C. Lamsdell, and Q. Liu. 2015. An unusual xipho- http://www.morphobank.org, accessed 2 June 2019. suran linking horseshoe crabs and eurypterids, from the Lower O’Reilly, J. E., M. N. Puttick, L. Parry, A. R. Tanner, J. E. Tarver, Devonian (Lochkovian) of Yunnan, China. Zoologica Scripta J. Fleming, D. Pisani, and P. C. J. Donoghue. 2016. Bayesian 44:645–652. methods outperform parsimony but at the expense of precision Shpinev, E. S., and D. V. Vasilenko. 2018. First fossil xiphosuran in the estimation of phylogeny from discrete morphological (Chelicerata, Xiphosura) egg clutch from the Carboniferous of data. Biology Letters 12:20160081. Khakassia. Paleontological Journal 52:400–404. Poty, E. 2010. Morphological limits to diversification of the rugose Shuster, C. N. Jr. 1982. A pictorial review of the natural history and and tabulate corals. Palaeoworld 19:389–400. ecology of the horseshoe crab, Limulus polyphemus, with reference Poulin, R. 2005. Evolutionary trends in body size of parasitic flat- to other Limulidae. Pp. 1–52 in J. Bonaventura, C. Bonaventura, worms. Biological Journal of the Linnean Society 85:181–189. and S. Tesh, eds. Physiology and biology of horseshoe crabs: stud- Prado, J. L., and M. T. Alberdi. 2008. Cladistic analysis among tri- ies on normal and environmentally stressed animals. Liss, lophodont gompotheres (Mammalia, Proboscidea) with special New York. attention to the South American genera. Palaeontology 51:903– Shuster, C. N. Jr., and K. Sekiguchi. 2003. Growing up takes about 915. ten years and eighteen stages. Pp. 103–132 in C. N. Shuster Jr., R. Puttick, M. N., J. E. O’Reilly, A. R. Tanner, J. F. Fleming, J. Clark, B. Barlow, and H. J. Brockmann, eds. The American horseshoe L. Holloway, J. Lozano-Fernandez, L. A. Parry, J. E. Tarver, crab. Harvard University Press, Cambridge, Mass. D. Pisani, and P. C. J. Donoghue. 2017. Uncertain-tree: discrimin- Simms, M. J. 1988. Patterns of evolution among Lower Jurassic cri- ating among competing approaches to the phylogenetic analysis noids. Historical Biology 1:17–44.

Simpson, C., and P. G. Harnik. 2009. Assessing the role of abun- Van Valkenburgh, B. 1991. Iterative evolution of hypercarnivory in dance in marine bivalve extinction over the post-Paleozoic. Paleo- canids (Mammalia, Carnivora)—evolutionary interactions biology 35:631–647. among sympatric predators. Paleobiology 17:340–362. Spearman, C. 1904. A proof and measurement of association Van Valkenburgh, B. 1999. Major patterns in the history of carniv- between two things. American Journal of Psychology 15:72– orous mammals. Annual Review of Earth and Planetary Sciences 101. 27:463–493. Stroud, J. T., and J. B. Losos. 2016. Ecological opportunity and Van Valkenburgh, B., X. Wang, and J. Damuth. 2004. Cope’s rule, adaptive radiation. Annual Review of Ecology, Evolution, and hypercarnivory, and extinction in North American canids. Sci- Systematics 47:507–532. ence 306:101–104. Tasch, P. 1961. Paleolimnology: Part 2—Harvey and Sedgwick Vrba, E. S., and N. Eldredge. 1984. Individuals, hierarchies and counties, Kansas: stratigraphy and biota. Journal of Paleontology processes: towards a more complete evolutionary theory. Paleo- 35:836–865. biology 10:146–171. Tashman, J. N., R. M. Feldmann, and C. E. Schweitzer. 2019. Mor- Wagner, P. J., and C. A. Sidor. 2000. Age rank/clade rank metrics– phological variation in the Pennsylvanian horseshoe crab sampling, taxonomy, and the meaning of “stratigraphic consist- Euproops danae (Meek & Worthen, 1865) (Xiphosurida, Euproopi- ency.” Systematic Biology 49:463–479. dae) from the lower Mercer Shale, Windber, Pennsylvania, USA. Watson, D. M. S. 1909. Limulus woodwardi, sp. nov., from the Lower Journal of Crustacean Biology 39:396–406. Oolite of England. Geological Magazine 6:14–16. Toljagić, O., and R. J. Butler. 2013. Triassic–Jurassic mass extinction Wright, A. M., and D. M. Hillis. 2014. Bayesian analysis using a as trigger for the Mesozoic radiation of crocodylomorphs. Biol- simple likelihood model outperforms parsimony for estimation ogy Letters 9:20130095. of phylogeny from discrete morphological data. PLoS ONE 9: Tucker, C. M., M. W. Cadotte, S. B. Carvalho, J. T. Davies, S. Ferrier, S. e109210. A.Fritz,R.Grenyer,M.R.Helmus,L.S.Jin,A.Ø.Mooers, Yoder, J. B., E. Clancey, S. Des Roches, J. M. Eastman, L. Gentry, S. Pavoine, O. Purschke, D. W. Redding, D. F. Rosauer, W. Godsoe, T. J. Hagey, D. Jochimsen, B. P. Oswald, M. Winter, and F. Mazel. 2017. A guide to phylogenetic metrics J. Robertson, B. A. J. Sarver, J. J. Schenks, S. F. Spear, and L. for conservation, community ecology and macroecology. Biological J. Harmon. 2010. Ecological opportunity and the origin of adap- Reviews 92:698–715 tive radiations. Journal of Evolutionary Biology 23:1581–1596.